AQL sampling is the quality bargain every serious bag program runs on: AQL sampling — acceptance quality limit, the statistical inspection system that pulls a defined random sample from a finished lot and accepts or rejects the whole lot on the defects found in it — exists because the two extremes both fail: inspecting nothing trusts the factory’s good intentions with your money, and inspecting everything costs more than the margin on most orders, so the industry standardized the middle path, a sample size and an accept-reject line that make quality a measurable, contractible event rather than a relationship hope. For waterproof bags the stakes double: the defects that matter most — a leaking seam, a mis-welded patch, a closure that almost seals — are exactly the ones the end customer discovers in the rain, after the freight, the duties and the launch budget are spent. This guide walks the system from the buyer’s chair: what AQL actually promises statistically, how the tables turn lot size into sample size, how defects are classified into critical, major and minor, why 2.5 became the industry default, what the system cannot catch, and the specification lines that make inspection binding on both sides of the purchase order.
Custom waterproof bag production since 2014, a 4,950 m² SGS-verified facility, 7 lines, 149 machines, 137 staff and 200,000 pieces monthly — that is the manufacturer reference at QUANZHOU JUNYUAN BAGS. Buyers work to MOQ 500 per design, samples 6–10 working days, bulk 35–50 days, AQL 2.5 inspection, 30/70 T/T, FOB Xiamen, shipping to 100+ countries.
The acceptance bargain: what AQL actually promises
AQL is widely quoted and widely misunderstood, so the honest starting point is statistical. An AQL plan does not promise that a passed lot is defect-free — it promises that a lot at the AQL defect level passes the inspection most of the time, and that lots substantially worse than the AQL fail most of the time. The number itself is a percentage: AQL 2.5 means the plan is calibrated around 2.5 defective pieces per hundred as the borderline quality level — lots at or below that level are accepted at high probability, lots well above it are rejected at high probability, and lots near it are where the sampling risk lives. Two risks ride every plan, and both are priced into the design: the buyer’s risk — a bad lot slipping through on a lucky sample — and the producer’s risk — a good lot failing on an unlucky one. The tables balance the two, and tightening the AQL shifts risk toward the producer while loosening it shifts risk toward the buyer. The practical translation for a first-time importer: AQL is a risk-sharing contract, not a quality guarantee — it makes quality statistically accountable across a long relationship of repeated lots, which is exactly how production business works. The full inspection architecture it plugs into — in-line checks, during-production inspection and the final random inspection this guide prices — is mapped end to end in our quality control guide, and the field document the inspector carries is the checklist our pre-shipment inspection guide details line by line.
The tables: from lot size to sample size
The code-letter and accept-reject tables buyers work from are the attribute-sampling tables published by the ISO as ISO 2859-1 and its national equivalents, and the statistics behind every sample size are set out in the NIST/SEMATECH engineering statistics handbook.
The AQL machinery runs on two tables, and reading them once removes the mystery permanently. Table one: the lot-size-to-code-letter table — the lot size (say 3,201 to 10,000 pieces) maps to a code letter under the chosen inspection level, and the letter is the key to everything downstream. The inspection level is the buyer’s dial: General Level II is the industry default — a balanced sample size; Level I samples less, cheaper and looser; Level III samples more, used when the product is failure-sensitive or the relationship is young. Special levels S-1 to S-4 exist for destructive or expensive tests — pulling a hydrostatic head or seam-burst check on a small sub-sample while the visual battery runs at Level II. Table two: the code-letter-to-plan table — the letter gives the sample size (for a 5,000-piece lot at Level II, 200 pieces) and, for each AQL value, the accept and reject numbers (at AQL 2.5 major on that sample: accept at 10 defectives, reject at 11). The discipline the tables enforce: the sample grows slower than the lot — a tenfold lot increase does not mean a tenfold sample — because statistics, not proportion, drives the confidence; and the accept number scales with the sample, so the same AQL is stricter in absolute terms on bigger lots. The buyer’s habit: never quote an AQL without its inspection level — "AQL 2.5 Level II" is a plan; "AQL 2.5" is a mood. And the randomness is the integrity: the inspector pulls cartons at random from across the lot, because a sample the factory selected is theater, not statistics.
Critical, major, minor: the defect taxonomy that drives the verdict
The verdict lives or dies on classification, and the three classes must be defined before the inspection starts — in the product specification, not in the argument afterward. Critical defects: anything that renders the product unsafe or unusable or violates regulation — for waterproof bags, the barrier failures: a leaking seam, a film breach, a closure that cannot seal, plus the universal safety set of sharp edges, broken needles, contaminated materials. Critical defects run at AQL 0 — not statistically zero risk, but the convention that one critical defect rejects the lot outright. Major defects: failures the end customer would notice and reject — wrong color against the approved swatch, broken stitching at a load point, a misaligned logo, a zipper that jams, a dimension outside tolerance. Major is the working battleground, conventionally set at AQL 2.5. Minor defects: cosmetic deviations most customers would not notice — a thread tail, a slight stitch irregularity off the face, a minor scuff on an interior surface — conventionally AQL 4.0. The waterproof-specific wrinkle: the classification table must name the leak tests — which seams get the water test, how many pieces, to what pressure — because a barrier failure discovered by the customer is the most expensive defect class in the category, and the only cheap place to catch it is the inspection table. The buyer’s discipline: write the defect list per product — twenty to forty named defects with their class — and attach it to the purchase order. An inspector without a defect list classifies by opinion, and opinion always drifts toward whoever is standing closest.
Why 2.5 became the industry default — and when to move it
AQL 2.5 for major defects is the soft-goods industry default not by law but by equilibrium: it is the tightest level most factories can hold economically across fashion-cycle variation, and the loosest level most brands can absorb in returns — the negotiated middle of fifty years of apparel and bag trade. Our own production standard runs AQL 2.5 as the stated frame, which is what the number means on our quotations: a calibrated, repeated, verifiable inspection event per lot. Moving the number is a real decision with real prices. Tightening to 1.5 or 1.0: appropriate for premium programs, licensed brands with retail-audit exposure, and products where a major defect is a safety-adjacent failure — waterproof medical and rescue builds live here; the price is more failed lots, more rework, and a higher unit cost, and the honest factory quotes it rather than pretending 1.0 costs the same as 2.5. Loosening to 4.0: appropriate for promotional price-point products where minor cosmetic variance is commercially irrelevant — trading inspection strictness for unit cost, a legitimate choice when named. The asymmetric truth buyers should hold: tightening the AQL does not improve the factory’s process — it improves the filtering of the output; the process improves through the upstream tools — in-line inspection, first-article discipline, the quality-management system our ISO 9001 guide prices — while AQL is the net at the dock. The strongest programs run both: a certified process making fewer defects, and a calibrated net catching the ones that remain. The number on the purchase order should be the number the product’s return budget can afford — decided once, in writing, before the first lot.
What the sample cannot see: the limits of AQL
AQL’s power is statistical, and its limits are statistical too — knowing them is what separates a quality system from a quality ritual. Rare-but-catastrophic defects: a failure present at one piece per thousand will be invisible to an 80- or 200-piece sample most of the time — which is why critical failure modes get 100% checks at the process step where they are cheap: every welded seam on a dry bag gets an air or water check on the line, because the dock sample cannot carry that load. Systematic defects: sampling catches random variation beautifully and systematic error badly — if the whole lot ran with the wrong thread color or a miscalibrated welder, the sample will show it, but only if the inspector tests the property that drifted; the defect list is the defense, and the during-production inspection is the early warning, the mid-process checkpoint our quality control guide positions. Deliberate concealment: AQL assumes an honest lot — the classic frauds (good cartons on top, golden samples for the inspector) are answered by the randomness protocol and, where risk justifies it, by container-loading supervision. Latent failures: the sample inspects the lot as it exists at the dock — it cannot see the seam that fails after six months of hydrolysis or the coating that chalks after a summer of UV, which is why the durability axes run as separate qualification programs, the aging discipline our accelerated aging guide standardizes. The summary: AQL is the last gate, not the whole fence — and buyers who understand its limits build the upstream system that makes the last gate a formality instead of a gamble.
Running the inspection: who, when, and with what leverage
The mechanics decide whether the AQL number is real. Who inspects: three models — the factory’s own QC (cheapest, structurally conflicted, acceptable only with a certified system and spot verification), the buyer’s own staff (highest control, highest cost, justified at scale), and third-party inspection companies (the industry standard: a per-man-day fee, a standardized report, and no stake in the verdict) — the lab-and-inspection landscape our third-party guide maps. When: the final random inspection fires at 100% packed or at least 80% packed — earlier, and the sample cannot represent the lot; later, and the leverage is gone. The leverage is the payment terms: the inspection sits between production and the balance payment, so a failed lot triggers rework and re-inspection before money moves — which is why the 30/70 structure with inspection as the 70-gate is the industry shape, and why buyers who pay in full before inspection have sold their AQL for nothing. The report: a professional inspection report carries the lot data, the sample pull record, defect photos with classification, the measurements, the on-site tests, and the accept-reject verdict against the named plan — a document that settles disputes because it was built to. Re-inspection: a failed lot gets sorted, reworked, and re-inspected on the full defective category — at the factory’s cost, which is the incentive alignment the whole system runs on. The buyer’s one-line audit: the AQL number protects you only if the inspection is random, the defect list is yours, the report is independent, and the money moves after the verdict — four conditions, all written into the purchase order.
The AQL purchase order: six lines
The guide compresses to six specification lines. One: the plan — AQL values per defect class (0 critical, 2.5 major, 4.0 minor as the working default) with the inspection level named, because a value without a level is a mood. Two: the defect list — twenty to forty named defects with classes, including the waterproof-specific leak tests and which seams they touch, attached to the PO. Three: the timing — final random inspection at 80–100% packed, with during-production inspection on lots above a defined size. Four: the inspector — named third-party or buyer staff, with the factory’s cooperation and carton access written in. Five: the leverage — balance payment gated on the passed report, rework and re-inspection at factory cost on failure. Six: the upstream system — the certified process, in-line checks and first-article discipline that make the dock verdict a formality, audited per the framework our BSCI audit guide and ISO 9001 guide detail. Six lines turn AQL from a brochure acronym into a binding quality contract — and the chain that executes it, 200,000 pieces monthly, samples in 6–10 working days, bulk 35–50 days, FOB Xiamen, is documented end to end for buyers who want the inspection system behind the number before the first lot.
Frequently Asked Questions
Q1. What is AQL sampling?
Acceptance quality limit: the statistical inspection system that pulls a defined random sample from a finished lot and accepts or rejects the whole lot on the defects found — the industry-standard middle path between inspecting nothing and inspecting everything.
Q2. What does AQL 2.5 mean?
The plan is calibrated around 2.5 defective pieces per hundred as the borderline level: lots at or below it pass with high probability, lots well above it fail with high probability. It is a risk-sharing contract, not a zero-defect promise.
Q3. What are the three defect classes?
Critical — unsafe, unusable or non-compliant, including barrier leaks, at AQL 0 (one rejects the lot); major — failures customers notice and reject, conventionally AQL 2.5; minor — cosmetic deviations, conventionally AQL 4.0.
Q4. How many pieces get inspected on a 5,000-piece lot?
At General Level II, a 200-piece sample — with accept at 10 major defectives and reject at 11 under AQL 2.5. The sample grows slower than the lot because statistics, not proportion, drives confidence.
Q5. What is General Inspection Level II?
The default sample-size dial: Level I samples less (cheaper, looser), Level III samples more (failure-sensitive products or young relationships). Special levels S-1 to S-4 serve destructive tests like hydrostatic head pulls.
Q6. When should the final inspection happen?
At 100% packed, or at minimum 80% packed — earlier and the sample cannot represent the lot; later and the payment leverage is gone. The inspection sits between production and the balance payment by design.
Q7. Can AQL catch every defect?
No — it is blind to rare-but-catastrophic defects (answered by 100% in-line checks at the failure step), latent durability failures (answered by aging qualification), and deliberate concealment (answered by randomness protocol and loading supervision).
Q8. Who should perform the inspection?
Third-party inspection companies are the industry standard — a per-man-day fee, a standardized report, no stake in the verdict. Factory QC works only with a certified system plus spot verification; buyer staff justifies its cost at scale.
Q9. Why is AQL 2.5 the industry default?
Equilibrium: the tightest level most factories hold economically and the loosest level most brands absorb in returns. Premium and safety-adjacent programs tighten to 1.5 or 1.0 and pay the rework price honestly.
Q10. Does tightening the AQL improve factory quality?
No — it improves filtering of output, not the process. Process quality improves through in-line inspection, first-article discipline and the certified quality system; AQL is the net at the dock, not the fence around it.
Q11. What is a critical defect in a waterproof bag?
Anything that kills the product promise or safety: leaking seams, film breaches, closures that cannot seal — plus sharps, broken needles, contamination. Critical defects reject the lot on the first instance.
Q12. What happens when a lot fails inspection?
The factory sorts and reworks the full defective category at its cost, then the lot re-inspects against the same plan. The balance payment holds until the re-inspection passes — the incentive alignment the system runs on.
Q13. Why does the defect list matter?
An inspector without a named defect list classifies by opinion, and opinion drifts toward whoever is standing closest. Twenty to forty named defects with classes, attached to the PO, make the verdict mechanical.
Q14. Is AQL 2.5 good enough for premium brands?
Usually not — premium, licensed and retail-audit-exposed programs run 1.5 or 1.0 on majors, accepting more rework and higher unit cost as the price of a tighter net. Decide once, in writing, before the first lot.
Q15. What is during-production inspection?
A mid-line checkpoint at 20–60% of production that catches systematic drift — wrong thread, miscalibrated welder — while correction is still cheap, complementing the final random inspection at the dock.
Q16. Do I need AQL inspection for a 500-piece order?
Yes — the tables scale down (an 80-piece sample at Level II), and the first order is exactly when the inspection earns most: it verifies the factory, not just the lot, and sets the precedent for every lot after.
Q17. What goes on an AQL specification?
Six lines: AQL values per class with inspection level, the named defect list with leak tests, timing at 80–100% packed, the named inspector, balance payment gated on the report, and the upstream certified process.
People Also Ask
What is a good AQL for bags?
The soft-goods default: 0 critical, 2.5 major, 4.0 minor at General Level II. Premium and safety-adjacent programs tighten majors to 1.5 or 1.0; promotional price-point builds may loosen minors — decided per return budget, in writing.
How is AQL inspection done?
The inspector pulls cartons randomly across the lot, samples per the code-letter table, checks each piece against the buyer’s defect list with measurements and on-site tests, classifies defects, and issues an accept-reject verdict against the named plan.
What does AQL 2.5 mean in inspection?
Lots at 2.5% defective pass with high probability; lots well above fail with high probability. On a 200-piece sample the accept-reject line sits at 10 versus 11 major defectives — a calibrated risk share, not a zero-defect guarantee.
What is the difference between major and minor defects?
Major: failures the end customer would notice and reject — wrong color, broken load-point stitching, jammed zipper, out-of-tolerance dimensions. Minor: cosmetic deviations most customers would not notice — thread tails, off-face stitch irregularity.
Why do inspections fail?
Most often systematic drift — a miscalibrated welder, a wrong material lot, an untrained line on a new style — surfacing as clustered majors. The defect list and during-production inspection catch drift early; the dock verdict is the last gate.
Can a factory cheat an AQL inspection?
The protocol is the defense: random carton pulls by an independent inspector, the buyer’s defect list, sealed samples, and loading supervision where risk justifies it. Cheating requires defeating the randomness — which is why the randomness is non-negotiable.